Toward Accurate Coastal Ocean Modeling
Peter C. ChuNaval Postgraduate SchoolMonterey, CA 93943, USA
Email: [email protected]://www.oc.nps.navy.mil/~chu
International Council for Sciences, Scientific Committee for Oceanic Research (SCOR)-, Miami, FL,
April 5-7, 2001.
Coastal Model
Lynch et al. (Oceanography 2001)
Major Problems in Coastal Modeling
• (1) Discretization• (2) Sigma Error• (3) Difference Schemes• (4) POM Capability • (5) Air-Ocean Coupling• (6) Severe Weather Effect • (7) Velocity Data Assimilation • (8) Turbulence/Wave Effects
(1) Discretization
Diversity in Discretization
• Finite Differences– Z – coordinate (MOM, …)– σ - coordinate (POM, COHERENS, etc…)– s- coordinate (SCRUM, ROMS …)– Layered/Isopycnal coordinates (NLOM,
MICOM, …)• Finite Elements
Z-Coordiante
• Note “staircase” topography representation, normally with no-slip conditions
Problems of the “Staircase Presentation”
• Difficult in simulating coastal flow.
• Example: Japan/East Sea (JES) Simulation (Kim and Yoon, 1998 JO)
JES Circulation Model Using MOM(Kim & Yoon, 1998)
• 1/6 deg resolution• 19 vertical level• Monthly mean wind
stress (Na et al. 1992)• Monthly mean heat
flux (Haney type)
Problem in Simulating Coastal Currents
• Model Observation
Layered/Isopycnal Coordinates• Pro
– Horizontal mixing is exactly along the surfaces of constant potential density
– Avoids inconsistencies between vertical and horizontal transport terms
• Con– It requires an evident
layered structure (not suitable for shelf circulation
– Some difficulty in modeling detrainment of ocean mixed layer
Layered/Isopycnal Coordinates
• (Metzger and Hurlburt1996, JGR)
• 1/8o, 6 layer with realistic bottom topography
• Not applicable to simulating shelf circulation
Sigma Coordinate Models
Sigma Coordinates
• Pro– Realistic Bottom
Topography
– Applicable to Shelf and Estuarine Circulation
• Con– Horizontal Pressure
gradient Error– High Vertical
Resolution in Shallow Water (Shelf) and Low Resolution in Deep Water
Horizontal Diffusion
• The second and fourth terms in the righthand side are neglected.
(2) Sigma Error
Pressure Gradient Error
Pressure Gradient Error
Seamount Test Case
Two Kinds of Sigma Errors(Mellor et al. 1998, JTECH)
• First Kind (SEFK):Horizontal Density GradientOscillatory Decaying
• Second Kind (SESF)– Vorticity Error
Reduction of Sigma Error
• Smoothing topography• Subtracting horizontally averaged density
field• Using generalized topography-following
coordinate system (e.g., S-coordinates in ROMS)
• Using high-order difference schemes
S-CoordinateGeneralized Topography-Following
Coordinates (Song & Haidvogel, 1994)
Error Analysis (S-Coordinate)
Error Evolution (S-coordinate)
• Radius of Seamount: r1 = 40 km, r2 =80 km
High-Order Schemes• Ordinary Five-Point Sixth-Order Scheme (Chu and Fan,
1997 JPO)• Three-Point Sixth-Order Combined Compact Difference
(CCD) Scheme (Chu and Fan, 1998 JCP)• Three-Point Sixth-Order Nonuniform CCD Scheme (Chu
and Fan, 1999, JCP)• Three-Point Sixth-Order Staggered CCD Scheme (Chu
and Fan, 2000, Math. & Comp. Modeling)• Accuracy Progressive Sixth-Order Scheme (Chu and
Fan, 2001, JTECH)
(3) Difference Schemes
Why do we need high-order schemes?
• (1) Most ocean circulation models are hydrostatic.
• (2) If keeping the same physics, the grid space (∆x) should be larger than certain criterion such that the aspect ratio
δ = H/ ∆x << 1
A Hidden Problem in Second Order Central Difference Scheme
• Both Φ’ and Φ’’ are not continuous at each grid point. This may cause some problems.
• Local HermitianPolynomials
Three-Point Sixth-Order Scheme
Three-Point Sixth Order CCD Schemes
• Existence of Global Hermitian Polynomials• First Derivative Continuous
• Second Derivative Continuous
Error Reduction Using CCD Schemes (Seamount)
Rotating Cone for Testing Various Schemes
Accuracy Comparison
(4) POM Capability
Chu et al 2001, JTECH
Evaluation of POM Using the South China Sea Monsoon Experiment (SCSMEX) Data
• IOP (April – June 1998)
T-S Diagram from SCSMEX Observations
Two Step Initialization of POM• (1) Spin-up
– Initial conditions: annual mean (T,S) + zero velocity– Climatological annual mean winds + Restoring type
thermohaline flux (2 years)• (2) Climatological Forcing
– Monthly mean winds + thermohaline fluxes from COADS (3 years) to 1 April
• (3) The final state of the previous step is the initial state of the following step
• (4) Synoptic Forcing– NCEP Winds and Fluxes: April 1 to June 30, 1998 (3 Months)
Two Types of Model Integration
• (1) MD1: – Without Data Assimilation – Hindcast Period: April-June 1998 (3 Months)
• (2) MD2: – With Daily SCSMEX-CTD Data Assimilation – Hindcast Period:
• May 1998: No data Assimilation in May • June 1998: No data Assimilation in June
Skill-Score
• Model-Data Difference
•• Mean Square Error
• Skill-Score (SS)
• SS > 0, Model has capability
Scatter Diagrams Between Model and Observation (MD1)
Histograms of (Model – Obs) for MD1
RMS Error for MD1 (No Assimilation)
Bias for MD1 (No Assimilation)
Skill-Score for MD1 (No Assimilation)
Scatter Diagrams for MD2 (with Assimilation)
RMS Error for MD2 (with Assimilation)
Bias for MD2 (with Assimilation)
Skill-Score for MD2 (with Assimilation)
Comments
• (1) POM-SCS has synoptic flux forcing.• (2) Without data assimilation, it has
capability to predict temperature, but not salinity.
• (3) With data assimilation, it has capability to predict salinity.
(5) Air-Ocean Coupling
• Coastal Atmosphere-Ocean Coupled System (CAOCS) for East Asian Marginal Sea (EAMS) Prediction
• Chu et al. (1999, JO)
Necessity for Air-Ocean Coupling
• (1) Sparse Meteorological Observation over Ocean
• (2) Uncertain Surface Fluxes
• (3) Nowcast/Forecast
Uncertain Atmospheric Forcing
RMS Difference Between NSCAT and NCEP Winds
Temporally Varying RMS Difference Between POM Model Results Under the Two Wind Forcing
(Chu et al. 1998, JGR)• Surface elevation
Temporally Varying RMS Difference Between POM Model Results Under the Two Wind Forcing
(Chu et al. 1998, JGR)• Velocity
Temporally Varying RMS Difference Between POM Model Results Under the Two Wind Forcing
(Chu et al. 1998, JGR)• Temperature
CAOCS Components
• Atmosphere: MM5-V3.4
• Ocean: POM
• Land Surface: BATS
CAOCS for East Asian Marginal Sea Prediction
Chu et al. (1999, 2000)
East Asian Circulation System
• Nitani (1972) Beardsley et al. (1983)
Area for Atmospheric Model
Distribution of Vegetation
Area for Ocean Model
Ocean Bottom
CAOCS Numerics• MM5V3.4
– Resolution• Horizontal: 30 km• Vertical: 16 Pressure Levels
– Time step: 2 min• POM
– Resolution• Horizontal: 1/6o × 1/6o
• Vertical: 23 σ levels– Time Steps: 25 s, 15 min
Ocean-Atmospheric Coupling
• Surface fluxes (excluding solar radiation) are of opposite signs and applied synchronously to MM5 and POM
• MM5 and POM Update fluxes every 15 min
• SST for MM5 is obtained from POM • Ocean wave effects (ongoing)
Lateral Boundary Conditions
• MM5: ECMWF T42
• POM: Lateral Transport at 142oE
MM5 Initialization
• Initialized from: 30 April 1998 (ECMWF T42)
Three-Step Initialization of POM• (1) Spin-up
– Initial conditions: annual mean (T,S) + zero velocity– Climatological annual mean winds + Restoring type
thermohaline flux (2 years)• (2) Climatological Forcing
– Monthly mean winds + thermohaline fluxes from COADS (3 years)
• (3) Synoptic Forcing– Winds and thermohaline fluxes from NCEP (1/1/96 – 4/30/98)
• (4) The final state of the previous step is the initial state of the following step
Simulated Surface Air Temperature, May 98
Volume Transport (Sv) Through Taiwan Strait
Volume Transport (Sv) Through Korean/Tsushima Strait
(6) Severe Weather Effect
• Response of the South China Sea to tropical cyclone Ernie 1996
• Chu et al. 2000 (JGR)
Tropical Cyclone Ernie 1996
Tropical Cyclone Wind Profile Model (Carr & Elsberry 1997, MWR)
• r ~ horizontal distance to the storm center• (uc, vc ) ~ radial and tangential velocities• γ ~ wind inflow angle to storm center• a = r/Rm , scaling factor • X ~ parameter
Surface Wind Field
• Vc ~ Wind field relative to the storm center (from the wind profile model)
• Vt ~ Strom translation velocity• Vbg ~ Back ground wind field
Computed Wind Field
NSCAT Winds
Time series of velocity and power density at 13oN and 119.5oE from
November 2 to 18, 1996
Comments
• Pom has a capability to simulate response of coastal water to tropical cyclones.
(7) Velocity Data Assimilation
Can we get the velocity signal from sparse and noisy data?
• Black Sea
Reconstruction of Velocity Field in Open Domain
Chu (2000)Chu and Ivanov (2001a,b)
Flow Decomposition
• 2 D Flow (Helmholtz)
• 3D Flow (Toroidal & Poloidal): Very popular in astrophysics
•
3D Incompressible Flow
•• When •u = 0• We have
Flow Decomposition
• 2 Ψ = - ζ, ζ is relative vorticity• 2Φ = - w
Boundary Conditions
Basis Functions
Flow Reconstruction
Reconstructed Circulation
Several Comments
• Reconstruction is a useful tool for processing real-time velocity data with short duration and limited-area sampling.
• The scheme can handle highly noisy data.• The scheme is model independent.• The scheme can be used for assimilating s
(8) Turbulence/Wave Effects
(a) Wave Momentum Flux in the Ocean (especially near the bottom)(b) Surface Roughness Length
(c) Wave Effect on TKE
Turbulence Parameterizations
• Bulk Mixed Layer Models – Garwood (1977), Price et al. (1986), Chu et al.
(1990), Chu and Garwood (1991), Chu (1993)
• Diffusion Models – Mellor and Yamada (1982)– Kantha and Clayson (1994)
Flow Decomposition
Momentum Flux
Turbulence/Wave Effects
Wave Stress in the Interior
Bottom Boundary Layer
Wave Effect on Surface Roughness Length
• TKΕ Dissipation in Air and Ocean are Functions of Wave Variables
• ε = f(h, λ, uph)
Wave Breaking & Turbulent Dissipation
• TKE Equation:
• D TKE/DT = S + WB ± B – D
– S: Shear Production– WB: Wave Breaking Effect– B: Buoyancy Production or Damping– D: Dissipation
• WB = γ (uphh)3/ λ4